This fix was proposed by Trond Myklebust. He says:
The type "sector_t" is heavily tied in to the block layer interface
as an offset/handle to a block, and is subject to a supposedly
block-specific configuration option: CONFIG_LBD. Despite this, it is
used in struct kstatfs to save a couple of bytes on the stack
whenever we call the filesystems' ->statfs().
So kstatfs's entries related to blocks are invalid on statfs64 for a
network filesystem which has more than 2^32-1 blocks when CONFIG_LBD
is disabled.
The content of the patch attached to this mail is below.
- struct kstatfs
Change the type of following entries from sector_t to u64.
f_blocks
f_bfree
f_bavail
f_files
f_ffree
Any feedback and comments are welcome.
Signed-off-by: Trond Myklebust <[email protected]>
Signed-off-by: Takashi Sato <[email protected]>
diff -uprN -X /home/sho/blocks/linux-2.6.15-rc5-blocks/Documentation/dontdiff linux-2.6.15-rc5-blocks/include/linux/statfs.h
linux-2.6.15-rc5-kstatfs/include/linux/statfs.h
--- linux-2.6.15-rc5-blocks/include/linux/statfs.h 2005-10-28 09:02:08.000000000 +0900
+++ linux-2.6.15-rc5-kstatfs/include/linux/statfs.h 2005-12-16 18:29:51.000000000 +0900
@@ -8,11 +8,11 @@
struct kstatfs {
long f_type;
long f_bsize;
- sector_t f_blocks;
- sector_t f_bfree;
- sector_t f_bavail;
- sector_t f_files;
- sector_t f_ffree;
+ u64 f_blocks;
+ u64 f_bfree;
+ u64 f_bavail;
+ u64 f_files;
+ u64 f_ffree;
__kernel_fsid_t f_fsid;
long f_namelen;
long f_frsize;
-- Takashi Sato
On Sat, 2005-12-17 at 03:39 +0800, Coywolf Qi Hunt wrote:
> That CONFIG_LBD is disabled means the kernel is not capable to attach
> a large block device, whether through network or locally attached. So
> in order to use a large network filesystem, simply enable LBD instead.
> Don't bother to bring unnecessary overheads.
It may surprise you to learn that not all network filesystems are block
based.
NFS has no truck with CONFIG_LBD at all.
Cheers,
Trond
On Sat, 2005-12-17 at 04:24 +0800, Coywolf Qi Hunt wrote:
> >
> > It may surprise you to learn that not all network filesystems are block
> > based.
> >
> > NFS has no truck with CONFIG_LBD at all.
>
> I thought no network filesystems are block based from a client
> viewpoint. (There's a network block driver though.) Client kernel
> needn't enable LBD.
Without this patch, the client _would_ have to enable LBD if it wanted
to correctly report the size of a large disk on the remote server.
The main point, though is that sector_t is a handle to a block. It is
_NOT_ the right type to use for reporting a disk size.
Cheers,
Trond